The rapid growth of AI technology has widened the gap between AI experts and the general population, making machine learning model creation largely inaccessible to non-experts. To bridge this divide, we introduce Agent for Machine Learning (A4ML), a fully automated text-to-model (T2M) system that converts natural language prediction objectives into accurate machine learning models. A4ML enhances accessibility through two key features: complete automation from data handling to model generation, and the production of comprehensive reports that include inference code, visualizations, and model analysis. Leveraging a multi-agent framework, A4ML manages the entire pipeline—from preprocessing data to generating interpretable outputs—requiring little technical expertise from the user. In evaluations across ten Kaggle datasets, A4ML successfully delivered fully automated models, achieving an average Normalized Mean Square Error of 0.2130 with hard-coded data analyser and 0.2220 with LLM-native counterpart, demonstrating its versatility and potential to democratize AI.